• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Wolfenstien 2 Vulkan benchmarks:

Nice matt, his post just reeks of hate because nvidia don't have this feature...

We continue enjoying higher res graphics than what nvidia can do :)
A little sore perhaps.

Shanks if you use that config, enable as much HBCC as you can.

If the game fails to start, your GPU does not have enough video memory. Not for the faint hearted.
 
FYI "2" Tb not 12 and GPUs used for 3D workloads like modeling can make use of the massive amount of available memory.
Real time video creation also benefits, did you not see the amd video editing real time 8k footage very nvidia? Let me tell you it was a one hose race.

HBCC isn't just for gaming FYI

And who in the right mind says no to anything that gives you the possibility to add more to a system?

https://www.tweaktown.com/news/58669/nvidia-beats-amd-real-time-8k-video-editing/index.html
 
Nice matt, his post just reeks of hate because nvidia don't have this feature...
We continue enjoying higher res graphics than what nvidia can do

It turned out that was not true. I looked at Matt config file and changed my file with both poolSize and r_shadowAtlasTileSize to see what happened with default 4096, 8192 and 16384. It was surprised ran and it worked on GTX 1070 after few attempts when it froze at press E to start game with both 8192 and 16384 by closed Edge browser with many tabs opened so I can enjoying same higher resolution graphics on 32GB DDR4 RAM as Matt. :D However I tried to take screenshots but all turned completed black so I took pictures with my Galaxy S6 instead.

4096

// This file is for local settings only and will not be cloud synchronized
configVersion 9
r_lightClassThreshold "3"
r_shadowsDistanceFadeMultiplier "2.5"
r_lightDistanceFadeMultiplier "2.5"
r_decalFilteringQuality "4"
r_decalLifetimeMultiplier "1.3"
r_decalDistanceFadeMultiplier "2"
r_clusteredDeferredRendering "1"
r_decalFadeCulling "0"
r_waterLodViewDistance "32"
r_waterInterleaveUpdates "0"
r_filmGrainRatio "4"
r_sharpening "4"
r_SSDOQuality "2"
r_particleFadeQualityMultiplier "2.5"
com_showConsumerPerfMetrics "2"
r_windowHeight "1080"
r_windowWidth "1920"
r_mode "19"
r_lodScaleDeviation "0"
r_swapInterval "0"
is_materialAniso "16"
is_poolSize "4096"
image_anisotropy "16"
r_userOptionSkipGPUCulling "0"
r_shadowAtlasTileSize "4096"
r_shadowAtlasHeight "16384"
r_shadowAtlasWidth "16384"
r_particlesLightAtlasHeight "4096"
r_particlesLightAtlasWidth "8192"
r_particlesLightAtlasQuality "3"

idbxZZP.jpg


161 FPS, used 7233MB VRAM and 7160MB 3200MHZ DDR4 RAM

8192

// This file is for local settings only and will not be cloud synchronized
configVersion 9
r_lightClassThreshold "3"
r_shadowsDistanceFadeMultiplier "2.5"
r_lightDistanceFadeMultiplier "2.5"
r_decalFilteringQuality "4"
r_decalLifetimeMultiplier "1.3"
r_decalDistanceFadeMultiplier "2"
r_clusteredDeferredRendering "1"
r_decalFadeCulling "0"
r_waterLodViewDistance "32"
r_waterInterleaveUpdates "0"
r_filmGrainRatio "4"
r_sharpening "4"
r_SSDOQuality "2"
r_particleFadeQualityMultiplier "2.5"
com_showConsumerPerfMetrics "2"
r_windowHeight "1080"
r_windowWidth "1920"
r_mode "19"
r_lodScaleDeviation "0"
r_swapInterval "0"
is_materialAniso "16"
is_poolSize "8192"
image_anisotropy "16"
r_userOptionSkipGPUCulling "0"
r_shadowAtlasTileSize "8192"
r_shadowAtlasHeight "16384"
r_shadowAtlasWidth "16384"
r_particlesLightAtlasHeight "4096"
r_particlesLightAtlasWidth "8192"
r_particlesLightAtlasQuality "3"

zmSblfE.jpg


125 FPS, used 7887MB VRAM and 10794MB 3200MHZ DDR4 RAM, surprised higher memory usage than at 4096 settings.

16384

// This file is for local settings only and will not be cloud synchronized
configVersion 9
r_lightClassThreshold "3"
r_shadowsDistanceFadeMultiplier "2.5"
r_lightDistanceFadeMultiplier "2.5"
r_decalFilteringQuality "4"
r_decalLifetimeMultiplier "1.3"
r_decalDistanceFadeMultiplier "2"
r_clusteredDeferredRendering "1"
r_decalFadeCulling "0"
r_waterLodViewDistance "32"
r_waterInterleaveUpdates "0"
r_filmGrainRatio "4"
r_sharpening "4"
r_SSDOQuality "2"
r_particleFadeQualityMultiplier "2.5"
com_showConsumerPerfMetrics "2"
r_windowHeight "1080"
r_windowWidth "1920"
r_mode "19"
r_lodScaleDeviation "0"
r_swapInterval "0"
is_materialAniso "16"
is_poolSize "16384"
image_anisotropy "16"
r_userOptionSkipGPUCulling "0"
r_shadowAtlasTileSize "16384"
r_shadowAtlasHeight "16384"
r_shadowAtlasWidth "16384"
r_particlesLightAtlasHeight "4096"
r_particlesLightAtlasWidth "8192"
r_particlesLightAtlasQuality "3"

RutoSH0.jpg


75 FPS, used 7894MB VRAM and 19059MB 3200MHZ DDR4 RAM, WOW massive memory usage than at 8192 settings and used more than 16GB memory. :eek: :cool:
 
Of course Nvidia have this feature with Compression and Tiled Caching.

It turned out that was not true. I looked at Matt config file and changed my file with both poolSize and r_shadowAtlasTileSize to see what happened with default 4096, 8192 and 16384. It was surprised ran and it worked on GTX 1070 after few attempts when it froze at press E to start game with both 8192 and 16384 by closed Edge browser with many tabs opened so I can enjoying same higher resolution graphics on 32GB DDR4 RAM as Matt. :D However I tried to take screenshots but all turned completed black so I took pictures with my Galaxy S6 instead.

4096

// This file is for local settings only and will not be cloud synchronized
configVersion 9
r_lightClassThreshold "3"
r_shadowsDistanceFadeMultiplier "2.5"
r_lightDistanceFadeMultiplier "2.5"
r_decalFilteringQuality "4"
r_decalLifetimeMultiplier "1.3"
r_decalDistanceFadeMultiplier "2"
r_clusteredDeferredRendering "1"
r_decalFadeCulling "0"
r_waterLodViewDistance "32"
r_waterInterleaveUpdates "0"
r_filmGrainRatio "4"
r_sharpening "4"
r_SSDOQuality "2"
r_particleFadeQualityMultiplier "2.5"
com_showConsumerPerfMetrics "2"
r_windowHeight "1080"
r_windowWidth "1920"
r_mode "19"
r_lodScaleDeviation "0"
r_swapInterval "0"
is_materialAniso "16"
is_poolSize "4096"
image_anisotropy "16"
r_userOptionSkipGPUCulling "0"
r_shadowAtlasTileSize "4096"
r_shadowAtlasHeight "16384"
r_shadowAtlasWidth "16384"
r_particlesLightAtlasHeight "4096"
r_particlesLightAtlasWidth "8192"
r_particlesLightAtlasQuality "3"

idbxZZP.jpg


161 FPS, used 7233MB VRAM and 7160MB 3200MHZ DDR4 RAM

8192

// This file is for local settings only and will not be cloud synchronized
configVersion 9
r_lightClassThreshold "3"
r_shadowsDistanceFadeMultiplier "2.5"
r_lightDistanceFadeMultiplier "2.5"
r_decalFilteringQuality "4"
r_decalLifetimeMultiplier "1.3"
r_decalDistanceFadeMultiplier "2"
r_clusteredDeferredRendering "1"
r_decalFadeCulling "0"
r_waterLodViewDistance "32"
r_waterInterleaveUpdates "0"
r_filmGrainRatio "4"
r_sharpening "4"
r_SSDOQuality "2"
r_particleFadeQualityMultiplier "2.5"
com_showConsumerPerfMetrics "2"
r_windowHeight "1080"
r_windowWidth "1920"
r_mode "19"
r_lodScaleDeviation "0"
r_swapInterval "0"
is_materialAniso "16"
is_poolSize "8192"
image_anisotropy "16"
r_userOptionSkipGPUCulling "0"
r_shadowAtlasTileSize "8192"
r_shadowAtlasHeight "16384"
r_shadowAtlasWidth "16384"
r_particlesLightAtlasHeight "4096"
r_particlesLightAtlasWidth "8192"
r_particlesLightAtlasQuality "3"

zmSblfE.jpg


125 FPS, used 7887MB VRAM and 10794MB 3200MHZ DDR4 RAM, surprised higher memory usage than at 4096 settings.

16384

// This file is for local settings only and will not be cloud synchronized
configVersion 9
r_lightClassThreshold "3"
r_shadowsDistanceFadeMultiplier "2.5"
r_lightDistanceFadeMultiplier "2.5"
r_decalFilteringQuality "4"
r_decalLifetimeMultiplier "1.3"
r_decalDistanceFadeMultiplier "2"
r_clusteredDeferredRendering "1"
r_decalFadeCulling "0"
r_waterLodViewDistance "32"
r_waterInterleaveUpdates "0"
r_filmGrainRatio "4"
r_sharpening "4"
r_SSDOQuality "2"
r_particleFadeQualityMultiplier "2.5"
com_showConsumerPerfMetrics "2"
r_windowHeight "1080"
r_windowWidth "1920"
r_mode "19"
r_lodScaleDeviation "0"
r_swapInterval "0"
is_materialAniso "16"
is_poolSize "16384"
image_anisotropy "16"
r_userOptionSkipGPUCulling "0"
r_shadowAtlasTileSize "16384"
r_shadowAtlasHeight "16384"
r_shadowAtlasWidth "16384"
r_particlesLightAtlasHeight "4096"
r_particlesLightAtlasWidth "8192"
r_particlesLightAtlasQuality "3"

RutoSH0.jpg


75 FPS, used 7894MB VRAM and 19059MB 3200MHZ DDR4 RAM, WOW massive memory usage than at 8192 settings and used more than 16GB memory. :eek: :cool:

Lol if you think HBCC is the same thing. Go and do your research for God sake.

Note AMDMatt VRAM not RAM, the reason you taking a big hit on FPS is because the game is now using out of Vram.
The performance drop of for a amd is much less and I sure @AMDMatt will show you this.
 
Lol if you think HBCC is the same thing. Go and do your research for God sake.

Note AMDMatt VRAM not RAM, the reason you taking a big hit on FPS is because the game is now using out of Vram.
The performance drop of for a amd is much less and I sure @AMDMatt will show you this.

nVidia have "pascal unified memory", etc. which is essentially the same as HBCC though AFAIK not possible currently to leverage with games unless they happen to be programmed for it (which I don't believe any are).

nVidia GPUs can page to system memory fine though HBCC can help to avoid some bottlenecks that the more primitive swap system will encounter - technically if you are properly employing HBCC you are also out of VRAM and ultimately still at the mercy of the bandwidth of the PCI-e interconnect unless you have additional local memory on the GPU.
 
nVidia have "pascal unified memory", etc. which is essentially the same as HBCC though AFAIK not possible currently to leverage with games unless they happen to be programmed for it (which I don't believe any are).

nVidia GPUs can page to system memory fine though HBCC can help to avoid some bottlenecks that the more primitive swap system will encounter - technically if you are properly employing HBCC you are also out of VRAM and ultimately still at the mercy of the bandwidth of the PCI-e interconnect unless you have additional local memory on the GPU.

The difference here is AMD you can set the amount of Available Memory for the Vram buffer to use. And they is hardly any performance hit! Unlike on Nvidia see his screenshots above the performance hit is massive and that's because all he doing is doing what any GPU does without HBCC and going to the much slower system RAM

With AMD HBCC its the cache controller here that helps reduce the performance bottleneck.. It works kinda like a Hybrid hard-drive in away that also have a slower drive boosted by a SSD in Cache.

Take note here of my GPUs dedicated Memory! This will not change like this on Nvidia unless they add a cache controller onto the GPU.
squad.jpg
 
Take note here of my GPUs dedicated Memory! This will not change like this on Nvidia unless they add a cache controller onto the GPU.

Pascal unified memory can do exactly the same thing without any additional hardware (nVidia just haven't bothered with it outside of CUDA as largely in the realworld its not particularly useful compared to developers actually implementing resource streaming/tiling systems that are optimised for their game). HBCC does allow you to make better use of the bandwidth available between the GPU and non-local or slower local memory and the bandwidth of that storage itself but it only goes so far - you can still saturate it and see the exact same performance slowdowns as the old paging system.
 
Pascal unified memory can do exactly the same thing without any additional hardware (nVidia just haven't bothered with it outside of CUDA as largely in the realworld its not particularly useful compared to developers actually implementing resource streaming/tiling systems that are optimised for their game). HBCC does allow you to make better use of the bandwidth available between the GPU and non-local or slower local memory and the bandwidth of that storage itself but it only goes so far - you can still saturate it and see the exact same performance slowdowns as the old paging system.

I hear that, and understand that it can do the samething. What I mean is Nvidia "User" is maybe what I should say can not set there own available memory pool to be used has the Dedicated Memory. This needs to be set by Nvidia themselves? Each game or something?

Surly they must have some additional hardware chip on the GPU? otherwise we back to square one and slow memory transfer speed between the GPU and Memory.

Not seen a single performance hit from using HBCC on AMD so far. if anything I have gained performance Squad for example seems to play much better. It has an option to load all textures into memory before game start this seem to help HBCC. i.e Am using 9GB Vram
 
The fine wine is in the accelerated ageing unit this time around.

Doing well for a compute card, no?
 
Looks like a toaster can run this game with ease lol. A game I will deffo be getting at some point though and enjoyed the last Wolfy game.
 
Looks like a toaster can run this game with ease lol. A game I will deffo be getting at some point though and enjoyed the last Wolfy game.

It's a very nice looking game in parts and have excellent effects. Highly recommend enabling the higher texture and shadow quality. 11Gb vram will be OK for 8k

This removes the lower res texture that sometimes can be seen.
 
It's a very nice looking game in parts and have excellent effects. Highly recommend enabling the higher texture and shadow quality. 11Gb vram will be OK for 8k

This removes the lower res texture that sometimes can be seen.

Thanks for the tip, gonna try this later/tomorrow :)

Hopefully my scrub of a system can manage 60fps+ ;)
 
Back
Top Bottom